Abstract:

In computer vision, image dehazing is an essential problem with applications ranging from surveillance systems to autonomous driving. In complicated settings with varied levels of haze intensity, traditional approaches frequently fail to remove haze efficiently. Deep learning techniques have demonstrated encouraging outcomes in tackling this issue in recent times. These techniques may efficiently describe the intricate links between hazy and haze-free images, and can train to recover clear images from hazy inputs by utilizing convolutional neural networks (CNNs) and generative adversarial networks (GANs). Recent developments in deep learning-based picture dehazing algorithms are examined in this abstract, together with their architectures, training approaches, and performance evaluation measures. Present issues and potential future directions in this quickly developing field are also highlighted. This project is mainly based on the dehazing the images using deep learning first get the input as the haze images and then pre-process it then apply splitting the data into the ratio of 8:2 and then apply deep learning algorithm for getting the ground truth images and then finally calculate the performance metrics such as MSE, RMSE are calculated for the retrieved images.

Description:

.